[ctypes-users] copying/slicing ctypes arrays, (c_ulong *n)()

[RJ <ra...@sa...>]

I have some code that gets a data array from a Win32 A/D dll/driver call at 250,000 samples/sec in 2K chunks. 
It has the form:
  buf = (c_ulong * 2000)()
which I then copy into a circular numarray buffer like:
  for i in range(2000): narray[pntr+i] = buf[i]
which is a bit slow, but fine for much lower data rates.

So, what is the fastest way copy the data?  I was re-reading the Python speed FAQ - I'll try 
  narray[pntr:pntr+2000] = map(None, buf)
tomorrow, but is that right?

Also, since narray is circular I sometimes need to split the ctypes buf in the copy when the pointer starts over, and ctypes does not seem to support slicing. So, I copy to a temp numarray and slice that, which seems sub-optimal.

The only "problem" here is that at 250Ksmps/sec it does not leave a lot of processor time to do other things.

Would it be best to use a large ctypes array instead of numarray as the circular buffer? I would still need to slice it though, with stepping, to do FFTs etc.

Thanks,
Ray


Secret anti-spam filter-passing text. Include with reply: 
qwertyuiop